Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but uniformly assigning a single response-level advantage to all tokens …
机构:上交
来源:arXiv 2610.32324 | AI4Papers 论文推荐平台